S UBMITTED TO T HE A STROPHYSICAL J OURNAL L ETTERS

Preprint typeset using LATEX style emulateapj v. 05/12/14

CHARACTERIZING INTRA-CLUSTER LIGHT AT z 0.5 IN THE HUBBLE FRONTIER FIELDS

TAKAHIRO M ORISHITA1,2,3 , L OUIS E. A BRAMSON1 , T OMMASO T REU1 , K ASPER B. S CHMIDT4 , B ENEDETTA V ULCANI5 , X IN WANG1

arXiv:1610.08503v1 [astro-ph.GA] 26 Oct 2016

Submitted to The Astrophysical Journal Letters

ABSTRACT
We investigate the intra-cluster light (ICL) in 4 Hubble Frontier Fields clusters at 0.3 < z < 0.6. Our new
method makes use of the unprecedented depth of Hubble Space Telescope imaging data to probe very diffuse
light (F160W 29.5 mag arcsec2 ) out to clustro-centric radii Rcl 300 kpc. The rest-frame B V color of
the ICL is consistent with that of passive cluster galaxies of log M /M < 9.5, 0.3 mag bluer than more
massive ones (log M /M > 10.5) and the cores of the brightest cluster galaxies, implying that the ICL and
more-massive cluster galaxies are built-up via distinct processes. The stellar mass of the ICL ranges from
11.1 < log MICL /M < 11.9, implying an ICL stellar mass fraction of 10-20%, about half of the local value.
Hence, we posit that the amount of ICL has rapidly increased since z 1, and is still being constructed, at a
rate of 200 M yr1 at z 0.5 by cluster specific mechanisms such as galaxy interactions and the stripping
of low-mass galaxies.
Keywords: galaxies: evolution galaxies: cluster galaxies: ICL
1. INTRODUCTION

Intra-cluster light (ICL) is starlight that fills the intergalactic space in dense galaxy environments. First proposed by
Zwicky (1937), it is unique to galaxy groups and clusters, suggesting that its formation process is related to environmentspecific phenomena that may also influence galaxies. Thus,
understanding the origin and evolution of the ICL may aid
our understanding of galaxy evolution.
Numerical calculations suggest that the ICL in massive
clusters (log M500 /M 15) formed at z < 1, epochs when
massive cluster galaxies (including brightest cluster galaxies;
BCGs) had nearly completed their stellar mass accumulation
(Murante et al. 2007; Collins et al. 2009; Contini et al. 2014;
Burke et al. 2015). Some fractions of stars in cluster member
galaxies remain in clusters after cluster specific mechanisms,
such as stellar/gas stripping and interaction between galaxies
and dark matter (Larson et al. 1980; Nipoti et al. 2003; McPartland et al. 2016). However, observational constraints on,
e.g., its origin and formation rate, have been limited largely
due to the low surface brightness of the ICL (e.g., Zaritsky
et al. 2004; Gonzalez et al. 2007; Guennou et al. 2012; Presotto et al. 2014, but see also Zibetti et al. 2005 who stack 683
SDSS clusters at 0.2 < z < 0.3), and difficulties introduced by
the fact that other cluster galaxies are, by definition, embedded within it.
In this letter, we utilize deep Hubble Space Telescope
(HST ) multi-band imaging from Hubble Frontier Fields
(HFF; Lotz et al. 2016) to overcome these obstacles, dissecting the ICL in four clusters at z 0.5when it is being
rapidly assembledout to Rcl 300 kpc. We develop a new
method to extract the ICL without any assumption on its light
mtaka@astro.ucla.edu
1 Department of Physics and Astronomy, University of California, Los Angeles, CA 90095-1547, USA
2 Astronomical Institute, Tohoku University, Aramaki, Aoba, Sendai 9808578, Japan
3 Institute for International Advanced Research and Education, Tohoku
University, Aramaki, Aoba, Sendai 980-8578, Japan
4 Leibniz-Institut fr Astrophysik Potsdam (AIP), An der Sternwarte 16,
D-14482 Potsdam, Germany
5 School of Physics, The University of Melbourne, VIC 3010, Australia

profile, relying only on much more robust knowledge of individual galaxy light profiles. By combining the deep data and
our new method, we can infer the ICLs:
1. rest-frame colors, and therefore its stellar population
and age;
2. total stellar mass, and therefore formation timescale, by
using comparisons with local clusters.
As we have a complete galaxy census (log M /M 7.8)
of each of these clusters (Morishita et al. 2016, hereafter
M16), we can also attempt to quantitatively link the ICL to
its probable cluster galaxy progenitors/sources.
Below, m = 0.3, = 0.7 and H0 = 70 km s1 Mpc1 , and
we assume a Chabrier (2003) initial mass function when estimating stellar masses. All magnitudes are quoted in the AB
system (Oke & Gunn 1983; Fukugita et al. 1996).
2. DATA AND METHOD

We base our analysis on the first four HFF clusters with

complete data: Abell2744, MACS0416, MACS0717, and
MACS1149. The HFF imaging spans ACS F435/606/814W
through WFC3IR F105/125/140/160W filters. The limiting
surface brightness (S/N > 3 per pixel) is F160W 29.5 mag
arcsec2 .
All ICL measurements are derived from the HFF cluster
central (CLS) pointing. The HFF parallel (PR1) pointings
lie at Rcl & 2 Mpc and therefore contain negligible ICL. We
use these data to estimate our background uncertainties (Section 2.1).
The method is schematically shown in Figure 1. We first
detect all the sources on the composite image using SExtractor (Bertin & Arnouts 1996). For galaxies with mF160W < 26
( 400-900 in each cluster), where structural fitting is reliable, we fit single Srsic light profiles using GALFIT (Peng
et al. 2002) in 300 300 pixel postage stamp (see M16).
At the same time, we fit the local sky background as a constant for each postage stamp. The estimated sky is dominated
by the ICL, but has also some contribution from non-ICL
component (e.g., zodiacal light). Although this is small in

The method is schematically shown in Figure 1. We first

detect all the sources on the composite image using SExtractor (Bertin & Arnouts 1996). For galaxies ( 400-900 per
each cluster) with mF160W < 26, where structural fitting is reliable, we fit single Srsic light profiles using GALFIT (Peng
et al. 2002) in 300 300 pixel postage stamp (see M16). At
the same time, we fit the local sky background as a constant
for each postage stamp.
2
M ORISHITA ET AL .
The estimated sky is dominated by the ICL, but has also
1. the ICLs rest-frame colors, and therefore its stellar popsome contribution from non-ICL component (e.g., earth zoulation and age;
Profile
fittingthe
in deep
a postage
diacal light).
Although
HST stamp;
imaging has lower sky
2. the ICLs total stellar mass,1and therefore formation
backgrounds compared to the ICL component in the IR bands,
timescale (using comparisons with local clusters).
this is not true
at bluer
wavelengths
where (ICL)
the ICL is faint (see
- Srsic
(galaxy)
+ Constant
below).
As we have a complete galaxy census (log M /M 7.8)4
After estimating the galaxy light profiles and local sky
300300
pixel Best
mtaka@astro.ucla.edu
skyi, and
i2 arewe
obtained
backgrounds
forfitallvalue,
the detected
sources,
then reconstruct
1 Department of Physics and Astronomy, University of California, Los
the ICL byforstacking
the modeled sky images. Since some
ith galaxy.
Angeles, CA 90095-1547, USA
2 Astronomical Institute, Tohoku University, Aramaki, Aoba, Sendai
postage stamps overlap, we calculate the representative value
980-8578, Japan
in each pixel,
by taking
weighted mean
ICLfICL
flux(x,iny),each
pixel the
is calculated
as; as
3 Institute for International Advanced Research and Education, Tohoku
PN
University, Aramaki, Aoba, Sendai 980-8578, Japan
2
skyi (x, y)/ i (x, y)
4 Available at http://glass.astro.ucla.edu
FICL (x, y) = i PN
,
(1)
2
i 1/ i (x, y)

secting the ICL in four clusters at z 0.5when it is being

rapidly assembledout to Rcl 300 kpc. We also develop
a new method to extract the ICL without any assumption on
its light profile, relying only on much more robust knowledge
of individual galaxy light profiles. By using the deep, Hubble Frontier Field (HFF; Lotz et al. 2016) data and our new
method, we can infer:

600

Number of pixel

500

ICL subtracted ( 2 =3.0)

Original ( 2 =4.5)
MACS0416

400

Repeat this for all galaxies in all

bands, to obtain the SED of ICLs.

2000

6
1500

5
4

1000

3
500

2
500

1000

1500

2000

2500

3000

4
3

2500

2
2000

1
0

1500

1
1000

2
3

500

4
0
0

500

1000

x (pixel)

1500

2000

2500

3000

300

200
100

y (pixel)

y (pixel)

2500

0
0

3000

count s 1 )

10

Number of stacking

3000

MACS0416CLS

Background (10

MACS0416CLS

0
0.005

x (pixel)

0.000
0.005
0.010
Background (count s 1 )

Figure 1. Left: Schematic view of ICL-subtraction method: 1. We fit the galaxy light profile and local sky background for each target galaxies in M16, and
2. stack the modeled sky in overlapped pixels and stack over the FoV of WFC3 ( 300 kpc) to construct the ICL map. 3. The stacking is weighted by the inverse
of 2 values of the light profile fitting (Eq. 1). Right: Comparison of two histograms for the pixel values in ICL-subtracted (red line) and original (black) images.
An asymmetric distribution (excess of positive pixel, i.e. ICL) is observed in the original one, while the distribution is more symmetric for the ICL-subtracted
image, as the medians of distributions indicate (bars of same colors as the distributions). The fit with a symmetric function for each distribution (dashed lines)
and its goodness (2 ) are shown.

the IR bands, this is not true at bluer wavelengths where the

ICL is faint (see below).
After estimating the galaxy light profiles and local sky
backgrounds for all the detected sources, we then reconstruct
the ICL by stacking the modeled sky images. Since some
postage stamps overlap, we calculate the representative value
in each pixel, FICL (x, y), by taking the weighted mean as
FICL (x, y) =

PN
i

skyi (x, y)/2i (x, y)

,
PN
2
i 1/i (x, y)

(1)

where skyi (x, y) is the best fit sky background in a given pixel
and 2i (x, y) is from GALFIT for the i-th galaxy. We repeat this analysis for all bands before fitting ICL SEDs (Section 2.2).
Although we have pixel-by-pixel information, we focus on
integrated ICL properties below. This is because systematic
uncertainties (Section 2.1) are non-negligible and will need to
be appropriately treated before, e.g., radial trends can be properly examined using the HFF or similar data (e.g., Presotto
et al. 2014; Montes & Trujillo 2014; DeMaio et al. 2015).
Figure 1 shows the histogram of pixel background values
in one of our clusters (MACS0416). Background pixels
are those without galaxy source light as defined using the
SExtractor segmentation map. The comparison between the
original and ICL-subtracted images clearly shows that the
former has an asymmetric tail to positive values while the
latter is more symmetric about zero. Fit with a symmetric
function probes this (2 = 4.5 and 3.0, respectively). We note
that the negative tail of ICL-subtracted image is similar to
the original imagesICL over-subtraction would lead to a
negative excess, which is not seen. This suggests that our ICL
subtraction (equivalently, ICL image construction; Figure 2)
is robust.

2.1. Uncertainties in the ICL Maps

Statistical uncertaintiesestimated from RMS maps made

publicly available by the HFF teamare very small given the
large numbers of pixels we are summing over.
In contrast, systematic uncertaintiesdominated by largerscale background fluctuations caused, e.g., by zodiacal light
and earth glow (Brammer et al. 2014)are significant.
Though they are partially suppressed by the HST reduction
pipeline,6,7 the residuals are sizable relative to the intrinsically faint signal from the ICL.
The effects of these residuals are estimated using the PR1
data, where the ICL contribution is negligible. We apply the
same method to the PR1 field to reconstruct an (ICL-free)
background map, and see the variance from the mean values as a function of bandpass. We adopt these values as our
ICL flux errors in the SED fitting procedure. They amount
to 30% in F160W and 50% in F606W.8 We ignore the
comparatively small statistical errors below.
2.2. Physical Parameter Estimates

The ICLs stellar mass and rest-frame colors are derived using FAST (Kriek et al. 2009). The input photometry is conducted on a common, resampled pixel grid for all filters. Pixels either on or near bright stars are masked. The systematic
errors estimated in Section 2.1 are used in the SED fitting.
Table 1 summarizes the results.
For consistency with the galaxy properties derived by M16,
we assume a Chabrier IMF and exponential star formation history for the ICL. The only difference here is our choice of a
6

http://www.stsci.edu/hst/campaigns/
frontier-fields/FF-Data
https://blogs.stsci.edu/newsletter/files/2015/03/
FFCalibration.pdf
8 F435W is dominated by the error, so we only quote upper-limits in that band.
Note that changing the size of the fitting postage stamps (from, e.g., 300 to
200 pixel) results in < 10% increases in F160W, which is much smaller than
the background uncertainty.
7

I NTRA - CLUSTER L IGHT IN THE HFF C LUSTERS

ABELL2744CLS

MACS0416CLS

B V (mag)

1.0

0.8

0.2Z
0.4Z
Z
2.5Z
ICL
BCG
Abell2744
MACS0416
MACS0717
MACS1149

3
2.0 Gyr
3.0 Gyr
5.0 Gyr
10.0 Gyr

AV = 1

0.6
Contour: galaxies of
log M /M (7.8, 9.0]

0.4

MACS0717CLS

MACS1149CLS

Figure 2.
RGB image of 4 ICL maps (blue:F435W, green:F125W,
red:F160W). The image size, 29002900 pixel, is comparable to the FoV
of WFC3.

0.4

0.6

0.8

1.0

1.2

1.4

(9.0, 9.5]
(9.5, 10.5]
(10.5, 12.0]

1.6

1.8

V J (mag)

dust-free SED model for the ICL and BCGs. This assumption
is reasonable for our purposes: there is no indication that the
ICL nor BCGs is very dusty (e.g., Kitayama et al. 2009).
Some studies report radial ICL metallicity gradients (e.g.,
Montes & Trujillo 2014), but because we work with spatially integrated (i.e., light-weighted) ICL data, we fixed this
quantity uniformly to a super-solar value (Z = 2.5 Z , where
Z = 0.02 is the solar metallicity). SED fitting results of ICL
and BCGs prefer super-solar to solar or sub-solar (Z = 0.4 Z )
metallicities: the bright metal-rich central ICL regions dominate over the faint less metal rich outskirts (e.g., DeMaio et al.
2015, see also Figure 3). Regardless, the effect of changing
the assumed metallicity from super-solar to solar causes its
stellar mass to increase only by 10%, which does not alter
our conclusions.
All quantities are derived similarly for the cluster BCGs for
comparison. Two BCGs are selected in each cluster by visual inspection. The location and number of the BCGs agree
with lens models9 in most of clusters. BCGs are known to
have diffuse envelopes extending to R > 100 kpc, and the stellar mass estimated by SExtractor measurements (MAG_AUTO;
see M16) is underestimated. We correct for this using the
GALFIT Srsic model magnitude.

Figure 3. Rest-frame colors of ICL (circles) in BVJ diagram. The dashed

and solid lines are grids for metallicity and ages, calculated with Bruzual
& Charlot (2003) model. The underlying contours are the colors of cluster
galaxies in four stellar mass bins, which are bounded at log M /M = 9.0, 9.5
and 10.5. The ICL colors are consistent with those of log M /M < 9.5
galaxies, while all BCGs reside the reddest region (stars). The effect by dust
attenuation (AV = 1.0 mag) on the grids is indicated with an arrow.

3. RESULTS

3.2. ICL Mass Fractions

The ICL mass fraction in clusters is an physically interesting quantity because it is a metric of how ICL formation proceeds with respect to cluster galaxy evolution. Figure 4 (and
Table 1) summarizes the total masses of the ICL component
(MICL ) and ICL+galaxies (Mtot. ) in each cluster system. The
total stellar mass is calculated in each cluster by summing the
masses of cluster galaxies with log M /M > 7.8 (see M16;
the contribution from lower-mass systems is negligible).
The stellar mass in ICL ranges in 11.1 < log MICL /M <
11.9similar to that of the BCGs. We note that our measurement for Abell 2744 (log MICL /M 11.18) is consistent with previous results derived from the same data used
here (Montes & Trujillo 2014).
The median ICL fraction, MICL /Mtot , is 1020% (Fig-

3.1. ICL Colors

Figure 3 shows the rest-frame colors of the ICLs of the four
HFF clusters. We use a BVJ color-color diagram to examine the age and metallicity of the ICLs stellar populations.
This avoids complexities associated with inferring rest-frame
U-band-dependent colors due to the poor S/N in the F435W
maps. The age-metallicity grid is based on a dust-free stellar
population model (Bruzual & Charlot 2003) consistent with
those used in the ICL/BCG SED fitting.
BCG colors are shown for comparison. These occupy
the right-top region of the diagram, implying more ma9

https://archive.stsci.edu/prepds/frontier/
lensmodels/

ture/older/more metal-rich stellar populations than those in

the ICL. That said, the BCGs in this study have color gradients of d(B V )/dR 0.006 mag kpc1 , implying a smooth
BCG/ICL transition at R 50 kpc.
Colors of normal cluster galaxies are taken from M16 and
shown in the background contours, grouped into three stellar
mass bins near log M /M = 9.0, 9.5, and 10.5. Since quiescent galaxies dominate in cluster cores, the blue cluster population is not clearly seen within the presented contour range
(see also M16 Figures 3 and 4). We see the consistency of
ICL and log M /M < 9.5 galaxies colors within errors.
Hence, taken at face value, Figure 3 implies that whatever
process built the ICL was distinct from that which built BCGs,
but may be related to the formation of comparatively lowmass (log M /M < 9.5) cluster galaxies. We return to this
point in Section 4.

M ORISHITA ET AL .
Table 1
Cluster and ICL properties.
Cluster

Redshift

rFoV
(Mpc)

(B V )ICL
(mag)

(V J)ICL
(mag)

MICL
(1011 M )

Mtot.
(1011 M )

M500
(1014 M )

FoV
M500
(1014 M )

MICL /Mtot.

/tot.

Abell2744
MACS0416
MACS0717
MACS1149

0.308
0.396
0.548
0.544

0.27
0.32
0.38
0.38

0.510.11
0.680.13
0.630.10
0.500.26

0.810.15
1.140.14
0.920.11
1.000.23

1.29+1.17
0.29
4.68+5.32
1.37
8.71+4.17
1.47
4.17+6.55
1.98

15.12+0.21
0.13
21.80+0.11
0.09
65.00+0.13
0.07
57.35+0.09
0.07

17.6 2.3
9.1 2.0
24.9 2.7
18.7 3.0

1.92 0.25
1.37 0.30
4.30 0.47
3.51 0.56

+0.08
0.080.02
+0.25
0.220.06
+0.06
0.130.02
+0.12
0.080.04

0.009 0.002
0.019 0.010
0.018 0.003
0.017 0.006

Note. Stellar mass is derived with the Chabrier IMF for galaxies, ICL, and BCGs. Dust extinction is set to be zero and metallicity is the super-solar value
(Z = 2.5Z ) for ICL and BCG.

Cumulative stellar fraction (log)

1
2

All
Red
Blue

Abell2744
+0.08
MICL /Mtot. = 0.080.02

4
0

BCGs
ICL
MACS0416
MICL /Mtot. = 0.22+0.25
0.06

1
2
3
4

MACS0717
+0.06
MICL /Mtot. = 0.130.02

10

11

12 8

MACS1149
MICL /Mtot. = 0.08+0.12
0.04

log M /M

10

11

12

Figure 4. Cumulative total stellar masses in 4 clusters. The total stellar

mass is the sum of star-forming (blue lines) and quiescent (red lines) cluster
galaxies, ICL (circles), and BCGs (stars). The values are normalized to stress
the fraction of ICL over the total stellar mass in each system, MICL /Mtot. .
The total stellar mass in the star-forming galaxies is deficit for the ICL mass
by itself, while passive systems have enough for ICL.

ure 4), which is about half of the local value (40%; Gonzalez et al. 2013), and consistent with the previous results at
similar redshifts (e.g., Burke et al. 2015), and theoretical expectations (e.g., Rudick et al. 2011; Contini et al. 2014). As
such, we can state that the ICL will double in mass between
z 0.5 and today. This is to be contrasted with the fact that at
z 0.5 BCG cores have essentially completed their formation
(e.g., Gallazzi et al. 2014), strengthening the distinct formation scenarios inferred from the BVJ diagram (Figure 3).
We show the cumulative stellar mass fractions in red and
blue galaxies in Figure 4. At the clustro-centric radii probed
by the CLS pointings, the total stellar mass in star forming cluster galaxies is < 10% of that in passive systems and
1050% of the ICLs (except in MACS1149, where they
are equal). In terms of ICL formation pathways, this would
suggest that the ICL may principally arise from the postprocessing of cluster passive galaxies, as opposed to the preprocessing of infalling blue galaxies (see Section 4).
The estimated total stellar fraction (ICL+BCGs+normal
galaxies, ) over the total mass (baryonic + dark matter,
tot. ) in the clusters is 1.8% except for Abell 2744 of 0.9%.

Total cluster masses are estimated by correcting the values of

M500 the mass within a radius where the clusters density exceeds 500 the critical density of the universefrom Mantz
et al. (2010) for our FoV. The measurements of total stellar
fraction, yet through an indirect method, are consistent at 1 with a measurement by Wang et al. (2015) and at 2- by
Hoag et al. (2016). We see no dependence of baryon fractions
on halo mass and redshift (cf. Gonzalez et al. 2007, who see
an anti-correlation).
4. DISCUSSION

We discuss possible scenarios for the origin and evolution

of the ICL based on the color and mass measurements described above and those of z 0 clusters.
Comparing our measurements to those of Gonzalez et al.
(2013) and computational predictions (Martel et al. 2012;
Contini et al. 2014), ICL mass fractions must approximately
double between z 0.5 and today. This implies growth rates
of 1012 M /5 Gyr = 200 M yr1 . This is a relatively high
absolute growth rate, but since these clusters harbor hundreds
to thousands of galaxies, it can be sustained from relatively
small leaks on a per-galaxy basis. The stars giving rise to
this mass growth of the ICL potentially comes from newly accreted galaxies, or the post-processing of already acquired
(likely passive) cluster galaxies.
Under the first scenario, since many of the newly accreted
galaxies will have been recently starforming (e.g., Dressler
et al. 2013), we might expect the ICL colors to remain relatively blue compared to the general, mature cluster galaxy
population. On the other hand, if the ICL was mainly grown
by post-processing, there would be no substantive blue population from which to draw ICL stars, so we would expect
passive evolution (of the older ICL component and the source
galaxies) to redden the observed ICL over time.
Studies of the Virgo clusterwhich is less massive and
therefore, if anything, younger than the descendants of
our clustersshow its ICL to have B V 0.8-1 (CastroRodriguz et al. 2009; Rudick et al. 2010, see also Zibetti
et al. 2005 for SDSS clusters at 0.2 < z < 0.3), about 0.2
mag redder than what we observe, though in consistent for
some of our clusters within error. Our calculation based on a
stellar population model (Galaxev; Bruzual & Charlot 2003)
shows the passive evolution of the observed population would
be consistent with the Virgo color. Hence, it would appear
that, at least at later times (z . 0.5), ICL growth is driven substantially by the removal of stars from the established cluster
population rather than infalling (at least recently starforming)
systems.
The origin of the ICL already established in the HFF sample is hazier. Regardless of the source of the stars, our findings suggest that the z 0.5 ICL is dominated by those with

I NTRA - CLUSTER L IGHT IN THE HFF C LUSTERS

ages 3-4 Gyr, implying its construction began at redshifts
of z . 1.4. This qualitatively concurs with Montes & Trujillo (2014)s observations of Abell2744, and numerical predictions by Contini et al. (2014).
Also in agreement with Montes & Trujillo (2014), the BVJ
diagram shows that the ICLs rest-frame colors to be more
consistent with those of log M /M < 9.5 galaxies than they
are with those of (at least the cores of) more massive galaxies
or BCGs, which are 0.3 mag redder in B V and therefore presumably much older. Given that lower-mass cluster
galaxies are also much more abundant than their higher-mass
counterparts (e.g., M16), this finding would argue for the ICL
to come more from the stripping/cannibalism (e.g., Ostriker
& Tremaine 1975; Moore et al. 1996; Treu et al. 2003; Nipoti
et al. 2003, 2004) of low-mass (log M /M . 10), as opposed
to even Milky Way mass galaxies. Given the results of M16
(see their Figures 9 and 11), perhaps half of these were already in the cluster at z 1, with the remainder having been
accreted subsequently, and so could indeed serve as a substantial ICL source population.
The interpretation is in opposition to that of Montes & Trujillo (2014) and predictions of Contini et al. (2014), which
suggest that Milky Way mass (log M /M & 10.5) progenitors are the most likely source for the ICL based on dynamical friction arguments. The presence of strong color gradients in such galaxies could remedy this discrepancy, since
the ICL is likely to come from easily stripped stars at large
galactocentric radii, which would also be the bluest (Figure 9
in Morishita et al. 2015). Additionally, the ICL mass in our
clusters is roughy equal to that of all extant log M /M < 10
red galaxies (Figure 4), implying that some help from highermass systems is likely (as favored by Montes & Trujillo
2014). This being said, it seems secure to posit from the HFF
data that the ICL does not arise through the same mechanisms
that establish at least the high-mass native cluster galaxy
population, including the BCG core. For example, dynamical
friction and the more massive galaxies can be responsible for
the ICL very close to BCGs, while the lower mass galaxies
and bluer ones are responsible for the rest of the ICL.
Hence, the ICL appears to have arisen at relatively late
times with early growth driven by a mix of the stripping of
preprocessed galaxies of log M /M . 10.5 and the extant
cluster population, and later growth (at z . 0.5) being dominated by the latter process (given the red ICL colors of z 0
clusters). Extremely deep imaging of z & 1 proto-clusters by
JWST and WFIRST will provide crucial leverage on the details of this history.
5. SUMMARY

1. We derived the SED properties (stellar mass and restframe colors) of ICL by using a new method, which is
consistent with the galaxies light profile fitting results.
2. Very deep imaging data allowed us to access the faint
stellar component out to 300 kpc from the cluster centers, 29.5 mag arcsec1 , and gave us more
accurate measurements of ICL for 4 HFF clusters at
0.3 < z < 0.6.
3. By comparing with BCGs and normal cluster galaxies,
we discussed the origin of ICL in terms of colors (i.e.
age) and total amount of the stellar mass. The ICL is
likely to rapidly form from quiescent cluster galaxies
of log M /M < 9.5 disrupted by stellar stripping since
z 1.4.

4. The stellar mass of the ICL is 10%-20% of that in

galaxies, about a half of the local value, suggesting that
the ICL is still under construction at the redshift observed. This also suggests that the ICL needs to be
constructed rapidly ( 200 M yr1 ) from z 0.5 to the
present date, emphasizing the importance of the clusterspecific mechanism on low-mass galaxies.
Support for this work is provided by NASA through HSTGO-13459. T.M. acknowledges support from the Japan Society for the Promotion of Science (JSPS) through JSPS research fellowships for Young Scientists.
REFERENCES
Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393
Brammer, G., Pirzkal, N., McCullough, P., & MacKenty, J. 2014,
Time-varying Excess Earth-glow Backgrounds in the WFC3/IR Channel,
Tech. rep.
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Burke, C., Hilton, M., & Collins, C. 2015, MNRAS, 449, 2353
Castro-Rodriguz, N., Arnaboldi, M., Aguerri, J. A. L., et al. 2009, A&A,
507, 621
Chabrier, G. 2003, PASP, 115, 763
Collins, C. A., Stott, J. P., Hilton, M., et al. 2009, Nature, 458, 603
Contini, E., De Lucia, G., Villalobos, ., & Borgani, S. 2014, MNRAS, 437,
3787
DeMaio, T., Gonzalez, A. H., Zabludoff, A., Zaritsky, D., & Bradac, M.
2015, MNRAS, 448, 1162
Dressler, A., Oemler, Jr., A., Poggianti, B. M., et al. 2013, ApJ, 770, 62
Fukugita, M., Ichikawa, T., Gunn, J. E., et al. 1996, AJ, 111, 1748
Gallazzi, A., Bell, E. F., Zibetti, S., Brinchmann, J., & Kelson, D. D. 2014,
ApJ, 788, 72
Gonzalez, A. H., Sivanandam, S., Zabludoff, A. I., & Zaritsky, D. 2013,
ApJ, 778, 14
Gonzalez, A. H., Zaritsky, D., & Zabludoff, A. I. 2007, ApJ, 666, 147
Guennou, L., Adami, C., Da Rocha, C., et al. 2012, A&A, 537, A64
Hoag, A., Huang, K.-H., Treu, T., et al. 2016, ArXiv e-prints,
arXiv:1603.00505
Kitayama, T., Ito, Y., Okada, Y., et al. 2009, ApJ, 695, 1191
Kriek, M., van Dokkum, P. G., Labb, I., et al. 2009, ApJ, 700, 221
Larson, R. B., Tinsley, B. M., & Caldwell, C. N. 1980, ApJ, 237, 692
Lotz, J. M., Koekemoer, A., Coe, D., et al. 2016, ArXiv e-prints,
arXiv:1605.06567
Mantz, A., Allen, S. W., Ebeling, H., Rapetti, D., & Drlica-Wagner, A. 2010,
MNRAS, 406, 1773
Martel, H., Barai, P., & Brito, W. 2012, ApJ, 757, 48
McPartland, C., Ebeling, H., Roediger, E., & Blumenthal, K. 2016,
MNRAS, 455, 2994
Montes, M., & Trujillo, I. 2014, ApJ, 794, 137
Moore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A. 1996, Nature,
379, 613
Morishita, T., Ichikawa, T., Noguchi, M., et al. 2015, ApJ, 805, 34
Morishita, T., Abramson, L. E., Treu, T., et al. 2016, ArXiv e-prints,
arXiv:1607.00384
Murante, G., Giovalli, M., Gerhard, O., et al. 2007, MNRAS, 377, 2
Nipoti, C., Stiavelli, M., Ciotti, L., Treu, T., & Rosati, P. 2003, MNRAS,
344, 748
Nipoti, C., Treu, T., Ciotti, L., & Stiavelli, M. 2004, MNRAS, 355, 1119
Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713
Ostriker, J. P., & Tremaine, S. D. 1975, ApJ, 202, L113
Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2002, AJ, 124, 266
Presotto, V., Girardi, M., Nonino, M., et al. 2014, A&A, 565, A126
Rudick, C. S., Mihos, J. C., Harding, P., et al. 2010, ApJ, 720, 569
Rudick, C. S., Mihos, J. C., & McBride, C. K. 2011, ApJ, 732, 48
Treu, T., Ellis, R. S., Kneib, J.-P., et al. 2003, ApJ, 591, 53
Wang, X., Hoag, A., Huang, K.-H., et al. 2015, ApJ, 811, 29
Zaritsky, D., Gonzalez, A. H., & Zabludoff, A. I. 2004, ApJ, 613, L93
Zibetti, S., White, S. D. M., Schneider, D. P., & Brinkmann, J. 2005,
MNRAS, 358, 949
Zwicky, F. 1937, ApJ, 86, 217